BigQuery is a warehouse-direct integration. Warehouse direct integrations authorize Predictable to directly query views and tables in the data warehouse that your organization manages.


To configure a BigQuery integration, your organization must complete the following steps:

Configure the Predictive Datasets

Create two datasets (PREDICT_INPUT and PREDICT_OUTPUT ) in a BigQuery project. This project will run the jobs for the SQL queries Predictable will execute, contain the data Predictable will query, and optionally receive the results of Predictable predictions.

Configure the Service Account

Create a Google Cloud Platform service account. This service account needs the following permissions:

  • At the Project level, in the project which contains the two datasets:
    • roles/bigquery.jobUser and roles/bigquery.readSessionUser consisting of:
      • bigquery.config.get
      • resourcemanager.projects.get
      • resourcemanager.projects.list
  • On the PREDICT_INPUT dataset:
    • roles/reader
  • On the PREDICT_OUTPUT dataset:
    • roles/bigquery.dataEditor
  • On the AUDIENCES dataset:
    • roles/bigquery.dataEditor

Once this service account is created, share the email address of that service account with your Predictable representative. In return, Predictable will provide a service account email address back. Delegate the above permissions by granting the token creator role to this Predictable-supplied service account.